Heterogeneity-agnostic AI/ML-assisted beam selection for multi-panel arrays

Ibrahim Kilinc; Robert W. Heath

eess.SPFeb 21, 2026

Heterogeneity-agnostic AI/ML-assisted beam selection for multi-panel arrays

Authors: Ibrahim Kilinc, Robert W. Heath

Organizations: Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093 USA

Abstract

AI/ML-based beam selection methods coupled with location information effectively reduce beam training overhead. Unfortunately, heterogeneous antenna hardware with varying dimensions, orientations, codebooks, element patterns, and polarization angles limits their feasibility and generalization. This challenge requires either a heterogeneity-agnostic model functional under these variations, or developing many models for each configuration, which is infeasible and expensive in practice. In this paper, we propose a unifying AI/ML-based beam selection algorithm supporting antenna heterogeneity by predicting wireless propagation characteristics independent of antenna configuration. We derive a reference signal received power (RSRP) model that decouples propagation characteristics from antenna configuration. We propose an optimization framework to extract propagation variables consisting of angle-of-arrival (AoA), angle-of-departure (AoD), and a matrix incorporating path gain and channel depolarization from beamformed RSRP measurements. We develop a three-stage autoregressive network to predict these variables from user location, enabling RSRP calculation and beam selection for arbitrary antenna configurations without retraining or having a separate model for each configuration. Simulation results show our heterogeneity-agnostic method provides spectral efficiency close to that of genie-aided selection both with and without antenna heterogeneity.

Figures & tables

Fig. 1: Global and local coordinate systems, and the communication system with a BS with single panel and a UE with multi-panel arrays. Antenna panels are UPAs with linearly polarized elements. Each panel has a 3D orientation vector based on the user orientation and the antenna placement.

	$\displaystyle\mathcal{R}^{(j)}=\argmin_{\{\hat{\mathbf{R}}_{\ell}\}_{\ell=1}^{L_{\text{all}}}}$	$\displaystyle f(\mathcal{A}^{(j)},\{\hat{\mathbf{R}}_{\ell}\}_{\ell=1}^{L_{\text{all}}})$
	$\displaystyle\text{s.t.}$	$\displaystyle\hat{\mathbf{R}}_{\ell}\succeq 0,$
		$\displaystyle\hat{\mathbf{R}}_{\ell}=\hat{\mathbf{R}}_{\ell}^{*}.$

Algorithm 1 Alternating Path Tracing Optimization

Fig. 2: Location-based three-stage path information predictor model. There are two common building blocks shown on the left. Dense layers with no activation specified have ReLU activation. The first stage predicts the AoA counts per angle cluster given the location information. Stage 2 performs an autoregressive AoD prediction for each AoA cluster index so that the context information avoids predicting the same AoD indices for the AoA index with more than one occurrence. Finally, stage 3 performs path matrix prediction given AoA, AoD one-hot vectors per path and the location.

Fig. 3: Urban canyon environment with dynamic vehicles. The buses are blockers and cars are users equipped with multi-panel antenna arrays. The solid bold red and green arrows show the orientation of the BS panel and a panel of a UE with multi-panels. Channels for each BS panel and UE panel pairs are generated in Sionna [ 34 ] . Different types of traced paths is visible in the legend.

Panel of Figure 5 — Fig. 4: Normalized signal power comparison between perfect CSI and traced paths at elevation angle $90^{\circ}$ . In (a), the BS beam is fixed to a DFT codebook beam; in (b), the UE beam is fixed. Traced paths from noisy RSRP provide angular power profiles matching perfect CSI.

Fig. 5: Mean SE vs beam coherence time $T_{\text{coh}}$ for $7\times 7$ UE panels. Shaded regions show Pareto frontiers for Static Pred and Traced Path with $N_{\text{b}}$ from 1 to 10 and solid lines use $N_{\text{b}}=6$ . Path tracing achieves within 0.5 b/s/Hz of the genie-aided method under antenna size heterogeneity.

Panel of Figure 7 — Fig. 6: Mean SE and user-for-user SE for orientation heterogeneity with $\Delta\alpha=96^{\circ}$ and $3\times 3$ UE panels. In (a), Pareto frontiers vary $N_{\text{b}}$ from 1 to 10. In (b), $T_{\text{coh}}=2$ ms. Static Pred and Traced Path achieve better user-for-user SE than Baseline under antenna orientation heterogeneity.

Fig. 7: Empirical CDF of SE for antenna size and orientation heterogeneity with $7\times 7$ panels, $\Delta\alpha=96^{\circ}$ , and $T_{\text{coh}}=3$ ms. Traced Path is within 1 b/s/Hz of Genie for over 99% of UEs and Static Pred achieves this for over 70% of UEs.

Fig. 8: Empirical CDF of SE for orientation and codebook heterogeneity with $3\times 3$ panels, $\Delta\alpha=96^{\circ}$ , $T_{\text{coh}}=2$ ms, and DFT test codebooks. Traced Path is within 1 b/s/Hz of Genie for over 99% of UEs and Static Pred for over 80%. Baseline performs poorly due to training on angle-limited codebooks.

Fig. 9: Mean SE vs $T_{\text{coh}}$ under antenna size, orientation, and codebook heterogeneity with $7\times 7$ panels, $\Delta\alpha=96^{\circ}$ , $T_{\text{coh}}=3$ ms, DFT codebooks, and $N_{\text{b}}=6$ . Our method has a potential to achieve within 0.5 b/s/Hz of the genie-aided method.

Wireless Communications

Wireless communications studies how information is sent over radio links and how networks manage signals, spectrum, and resources reliably.

See all 399 papers

Rethinking Beam Management: Generalization Limits Under Hardware Heterogeneity
Feb 20, 2026Nikita Zeulin, Olga Galinina, Ibrahim Kilinc +2Wireless Communications OOD Generalization
Tampere University, Tampere, Finland · University of California San Diego, CA, USA
BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management
Sep 1, 2026Yue Zhang, Xiucheng Wang, Wenshuo Chen +16G Networks Radio Propagation Modeling
State Key Laboratory of ISN and the School of Telecommunications Engineering, Xidian University, Xi’an 710071, China · The Hong Kong University of Science and Technology (Guangzhou), Guangzhou 511453, China
Bypassing the CSI Bottleneck: MARL-Driven Spatial Control for Reflector Arrays
Apr 6, 2026Hieu Le, Oguz Bedir, Mostafa Ibrahim +2Wireless Communications Reconfigurable Intelligent Surfaces
Electrical and Computer Engineering, Texas A&M University, College Station, Texas, USA · Engineering Technology and Industrial Distribution, Texas A&M University, College Station, Texas, USA · School of Performance, Visualization, and Fine Arts, Texas A&M University, College Station, Texas, USA +1

Feb 20, 2026J/K move · Enter open · S save

Rethinking Beam Management: Generalization Limits Under Hardware Heterogeneity

Nikita Zeulin, Olga Galinina, Ibrahim Kilinc, Sergey Andreev+1

Tampere University, Tampere, Finland · University of California San Diego, CA, USA

Wireless Communications OOD Generalization Wireless Resource Allocation

Open paper AskPDF arXiv

Hardware heterogeneity across diverse user devices poses new challenges for beam-based communication in 5G and beyond. This heterogeneity limits the applicability of machine learning (ML)-based algorithms. This article highlights the critical need to treat hardware heterogeneity as a first-class design concern in ML-aided beam management. We analyze key failure modes in the presence of heterogeneity and present case studies demonstrating their performance impact. Finally, we discuss potential strategies to improve generalization in beam management.

Abstract

Wireless Communications

Explore similar work

Rethinking Beam Management: Generalization Limits Under Hardware Heterogeneity

BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management

Bypassing the CSI Bottleneck: MARL-Driven Spatial Control for Reflector Arrays